Prediction of Stock Trading Signal Using Colored Petri Nets and Genetic Algorithm (Case study: Tehran Stock Exchange)

Deciding when to buy or sell stocks is a challenging problem for investors to increase incoming and decrease loss in stock market. Methods of predicting stock market in literature fall into two main categories: fundamental-analysis-based methods and technical-analysis-based methods. Predicting the trend of stock price movements and detecting changes of trend direction using technical analysis is generally preferred by analyzers in comparison with price prediction methods using fundamental analysis, due to data frequency reduction and less data variations in short-term. Most of the both methods use Artificial Intelligence (AI) techniques such as data mining and meta-heuristic approaches. AI based approaches suffer disadvantage of expert interaction necessity. In this paper a hybrid method of Genetic Algorithm (GA) and Colored Petri Nets (CPN) is proposed to model, simulate and predict buy/sell stock trading signals. CPN is a formal modeling language which supports mathematical simulation and markup language programming that reduces the necessity of human expert interaction in prediction approach. Stock trading rules are extracted from historical data of 162 companies in Tehran stock exchange weekly gathered from April 2016 till April 2018 using GA to maximize earning per share. Simulation results demonstrate that proposed method outperforms other state-of-the-art methods, in terms of classification correctness.